Further observations on periodicities of nucleotide occurrences in natural DNA's.

Abstract

There are non-random features in the occurrences of nucleotides in the DNA's of certain organisms which are detectable by statistical analyses of the entire sequence. Earlier, using the bacteriophage Phi-X 174 DNA sequence, we had reported that the self-information values for one type of dinucleotide association showed a marked periodicity when their autocorrelation coefficients were graphed. A similar, but computationally simpler, analysis has been developed which gives a comparable indication of periodicity. The difference, in average autocorrelation coefficients obtained with this analysis, between the peak values and all others has been used as an index to compare the extent of periodic non-randomness for a series of natural DNA sequences and for various artificial sequences. Calculations show that triplet periodicity, the relationship between dinucleotides separated by a single nucleotide, is characteristic only of the natural sequences of certain filamentous phages and is not found prominently in any other DNA analyzed (including sequences of similar length from plasmids, yeast, bacteria and higher animals). By shuffling nucleotides in a given sequence or by substituting selected nucleotides to alter various positions in both periodic and aperiodic sequences, we have found that an excess or deficiency of a given nucleotide at one of the three positions in a triplet reading frame can simulate the periodic characteristic. Thus, it appears that this global statistical analysis detects the tendency for single-strand phages to utilize a specific nucleotide, rather than one randomly selected, to constitute codons.